Accelerate your journey of protein discovery

Our technology enables optimization of proteins and RNAs for desired properties including binding affinity, stability, escape resistance, and low immunogenicity.

Step 1

Design Multi-objective Fitness Function

Define desired properties and constraints

We work with your team to define a project-specific fitness function integrating evolutionary constraints, binding affinity, stability, expression, and manufacturability. Benchmarks are performed on public data to validate performance.

Literature → Constraints → Fitness Function
Design Multi-objective Fitness Function

Define desired properties and constraints

Literature → Constraints → Fitness Function

Step 2

Few-shot Prediction & Experimental Testing

Select optimal variants to test

Our platform generates high-priority candidates using few-shot prediction models. Selected variants are tested experimentally — either in your lab or through CROs — to validate the fitness landscape and guide optimization.

Prediction → Selection → Wet-lab testing
Few-shot Prediction & Experimental Testing

Select optimal variants to test

Prediction → Selection → Wet-lab testing

Step 3

Retrain & Optimize

Iterative learning with experimental feedback

We retrain the machine learning models on experimental data to refine predictions and generate new, optimized batches of candidates. This iterative DBTL cycle converges rapidly towards highly fit, robust protein designs.

Retrain → Next-generation design → DBTL Cycle
Retrain & Optimize

Iterative learning with experimental feedback

Retrain → Next-generation design → DBTL Cycle